(1) Background: This research investigates the application of Artificial Intelligence (AI), particularly biomimetic convolutional neural networks (CNNs), for the automatic classification of gastrointestinal (GI) polyps in endoscopic images. The study combines AI and Transfer learning techniques to support early detection of colorectal cancer by enhancing diagnostic accuracy with pre-trained models; (2) Methods: The Kvasir dataset, comprising 4000 annotated endoscopic images across eight polyp categories, was used. Images were pre-processed via normalisation, resizing, and data augmentation. Several CNN architectures, including state-of-the-art optimized ResNet50, DenseNet121, and MobileNetV2, were trained and evaluated. Models were assessed through training, validation, and testing phases, using performance metrics such as overall accuracy, confusion matrix, precision, recall, and F1 score; (3) Results: ResNet50 achieved the highest validation accuracy at 90.5%, followed closely by DenseNet121 with 87.5% and MobileNetV2 with 86.5%. The models demonstrated good generalisation, with small differences between training and validation accuracy. The average inference time was under 0.5 s on a computer with limited resources, confirming real-time applicability. Confusion matrix analysis indicates that common errors frequently occur between visually similar classes, particularly when reviewed by less-experienced medical physicians. These errors underscore the difficulty of distinguishing subtle features in gastrointestinal imagery and highlight the value of model-assisted diagnostics; (4) Conclusions: The obtained results confirm that Deep learning-based CNN architectures, combined with Transfer learning and optimisation techniques, can classify accurately endoscopic images and support medical diagnostics.

Introduction: Classifying gastrointestinal (GI) polyps detected in colonoscopy images is a critical task in colorectal cancer prevention. Given the diagnostic ambiguity of serrated polyps, which share morphological features with both hyperplastic and adenomatous lesions, this study focuses on multiclass classification using machine learning (ML) techniques. Multiclass Logistic Regression (LR), a model favored by clinicians for its interpretability, was initially optimized and evaluated. Methods: A structured dataset comprising 152 instances and 698 extracted features was used. We conducted a statistical analysis of 88 LR configurations, varying solvers, penalties, and regularization strengths. To improve classification performance, four additional ML algorithms were implemented: k-Nearest Neighbors (kNN), Support Vector Machine (SVM), Random Forest (RF), and XGBoost. For each classifier, parameter tuning was applied using grid search and stratified cross-validation. Results: The best-performing LR model (liblinear solver, L1 penalty, C = 0.01) achieved an accuracy of 70.39%, outperforming physician benchmarks (experts: 65.00%, beginners: 58.42%). In the multiclass setting, XGBoost achieved the highest macro-average F1-score (0.88) and overall accuracy (89.34%), followed by Random Forest (F1 = 0.85, accuracy = 86.05%), SVM (F1 = 0.83, accuracy = 84.21%), and kNN (F1 = 0.56, accuracy = 66.38%). Discussion: While LR remains valuable for its interpretability, ensemble methods such as XGBoost and Random Forest demonstrated superior performance and robustness. These findings support the integration of advanced ML models into clinical decision support systems, particularly in low-data scenarios where deep learning may be impractical.

In this study, we used the hydrophobic-polar (HP) two-dimensional square and three-dimensional cubic lattice models for the problem of protein structure prediction (PSP). This kind of lattice reduces computational time and calculations, the conformational space from $9^n$ to $3^{n-2}$ for the 2D square lattice and $5^{n-2}$ for the 3D cubic lattice. Even within this context, it remains challenging for genetic algorithms or other metaheuristics to identify the optimal solutions. The contributions of the paper consist of: (1) implementation of a high-performing novel genetic algorithm (GA); instead of considering only the self-avoiding walk (SAW) conformations approached in other work, we decided to allow any conformation to appear in the population at all stages of the proposed all conformations biomimetic genetic algorithm (ACGA). This increases the probability of achieving good conformations (self avoiding walk ones), with the lowest energy. In addition to classical crossover and mutation operators, (2) we introduced specific translation operators for these two operations. We have proposed and implemented an HP Protein Visualizer tool which offers interpretability, a hybrid approach in that the visualizer gives some insight to the algorithm, that analyse and optimise protein structures HP model. The program resulted based on performed research, provides a molecular modeling tool for studying protein folding using technologies such as Node.js, Express and p5js for 3D rendering, and includes optimization algorithms to simulate protein folding.

Given the complexity and variability of gastrointestinal polyps, there is a growing need for more accurate machine-learning algorithms that can assist medical professionals in decision-making processes. In our study, we employed machine learning (ML) techniques to classify gastrointestinal polyps seen in colonoscopy images, that present a risk of colon cancer, affecting a large portion of the population. The dataset used in the research comprised 152 instances and included three types of lesions, totaling 76 polyps. The study consisted of applying Logistic Regression (LR) to classify gastrointestinal images. The principal motivation for choosing this algorithm consists in the fact that is preferred by many medical researchers. Another motivation consisted in the fact to compare the polyps classification accuracy of physicians and LR with optimized hyperparameters. The approached model performance was evaluated with the following metrics: Accuracy, Precision, Recall, F1_Score and Macro-average. We used the Multiclass Logistic Regression (mLR) classifier to classify polyps into hyperplastic (27.63%), serrated (19.74%), and adenoma lesions (52.63%). It should be noted that serrated polyps are hard to diagnose, even by physicians, because they have characteristics of both hyperplastic and adenoma ones. One of the objectives consisted in the study of the hyperparameters tunning of the Logistic Regression to find the best-fitted model. To check the influence of the data mix on the results obtained, for the two best models (LR with liblinear solver, L1 penalty and C = 0.01), it was performed a comprehensive statistical analysis. Based on an algorithm that admits obtaining accurate results. Such analyses are mistakenly applied in the scientific literature. The model with optimized hyperparameters achieved an accuracy score of 70.39%, indicating a reasonable level of prediction, surpassing both beginner and expert physicians and obtaining results similar to those obtained by Random Subspace and Random Forest. The model successfully classifies the hyperplastic and adenoma polyps but showed to have a more limited ability to predict the serrated ones.